Skip to main content

Cluster Map

The Cluster Map page (nav label Health Map, /kubernetes/map, under the Infrastructure group) is a plain-language health map of a single cluster — it shows at a glance where a problem is and what it affects, without needing to read raw Kubernetes objects. It pairs a written "needs attention" summary with an interactive topology graph.

Cluster and namespace selectors​

Two selectors at the top scope the map: Cluster picks one registered cluster, and Namespace narrows to a single namespace or All namespaces.

Needs attention​

A highlighted panel lists the current problems in plain English rather than status codes — for example, impact.degraded — the app's own copies are failing — the machines they run on are healthy, with a Why: line naming the underlying cause (e.g. ContainerCreating · pod payments-service-…). The count in the panel header tells you how many issues are open, and each entry expands to the affected object.

The map​

Below the panel is an interactive graph. The Show machines toggle adds the underlying nodes, and a legend marks each object Healthy, Affected or Fault. The graph drills through the cluster's structure:

  • Cluster — The whole cluster, with its rolled-up health.
  • App group — A namespace, shown as a group of applications.
  • Application — A workload, with its running-replica count and an "app problem" badge when impaired.
  • Instance — An individual pod, with its phase (e.g. ContainerCreating).
  • Nodes — The machines the workloads run on, so you can tell an application fault from a machine fault.

Following the chain from a red application down to a healthy machine is what makes the "the app's copies are failing, the machines are fine" story obvious at a glance.

Best Practices​

  • Read the Needs attention panel first, then use the map to confirm the blast radius — the panel tells you what, the graph tells you how far.
  • Turn on Show machines when an application looks unhealthy but you can't tell why; a shared node fault often explains several sick applications at once.
  • Narrow the Namespace selector on a busy cluster — the all-namespaces view is comprehensive but dense when you already know which team is affected.
  • Kubernetes — Cluster registration and per-cluster resource tabs
  • Workloads — The ranked, cross-cluster health table
  • K8s Intelligence — Current, past and predicted Kubernetes problems